The farming robots that will feed the world | Hard Reset

Summary

The video script introduces a revolutionary robot farmer developed by Dogtooth, a company co-founded by Duncan, who has a background in machine learning and computer vision. The robot, designed to address labor shortages in the agricultural sector, is capable of delicately picking soft fruit like strawberries, a task previously only performed by humans. The U.K., facing a significant labor shortage for fruit picking, resulting in substantial fruit waste and economic loss, stands to benefit greatly from this technology. The robot's sophisticated arm, custom-designed for the task, uses stereo cameras and computer vision to navigate the farm environment, identify, and harvest the fruit. The technology not only picks fruit but also inspects it for quality, sorting it into waste or retail containers. While the current generation of robots works slower than humans, they are designed for endurance, capable of operating through multiple shifts without fatigue. The robots are also mobile, with some being shipped to Tasmania for seasonal work, highlighting the potential for global agricultural support. The script discusses the broader implications of robotic automation on labor markets and the environment, suggesting that if managed correctly, this technology could create more desirable jobs for humans while reducing the carbon footprint of food production.

Takeaways

• 🤖 **Robotic Innovation in Farming**: The development of robots capable of delicate tasks like picking soft fruit is a significant advancement in agricultural automation.
• 🇬🇧 **UK's Agricultural Challenge**: There is a shortage of human labor for fruit picking in the UK, leading to significant waste and economic loss.
• 🍓 **Precision Picking**: Robotic systems are designed to be smart and dexterous to pick strawberries without damaging them, which is crucial for fruit quality.
• 🚜 **Beyond Tractors**: While farms already use automated technologies like GPS-guided tractors, the new robots represent a leap to handling more complex and delicate tasks.
• 🤓 **Interdisciplinary Founders**: The co-founders of Dogtooth come from diverse backgrounds, including machine learning and computer vision, which is instrumental in creating the robots.
• 🧼 **Custom Design Necessity**: Due to the unique demands of farming, most components of the robots, including the arms, had to be custom designed and built from scratch.
• 📈 **Data-Driven Agriculture**: The robots harvest not only fruit but also data, which can be used to improve yield prediction and farming practices.
• 🌱 **Seasonal Workforce Supplement**: The aim is not to replace human workers but to assist in tasks that are difficult to staff, such as seasonal fruit picking.
• 🌍 **Global Workforce**: These robots are designed to be mobile and work in different environments worldwide, even spending parts of the year in Australia.
• 🔄 **Environmental Impact**: By enabling local production of crops, the use of robots could reduce the carbon footprint associated with transporting produce over long distances.
• ⚙️ **Future of Work**: The integration of robotics in labor raises questions about the future of human jobs, emphasizing the need to find a balance between automation and human employment.

Q & A

What is the primary function of the robot farmer developed by Dogtooth?

The primary function of the robot farmer is to pick soft fruit like strawberries, which is a task that requires a high level of dexterity and precision.

Why is it challenging to find human labor for harvesting strawberries in the U.K.?

It is challenging to find human labor for harvesting strawberries in the U.K. due to a labor shortage, which has led to significant fruit wastage and financial loss for the farming industry.

How does the robot's arm work in terms of positioning and movement?

The robot's arm can move to any position in 3D space and approach the target from any angle, allowing it to handle delicate tasks such as picking strawberries without damaging them.

What is the significance of the cameras and LED lights in the robot's pickerhead?

The cameras in the pickerhead allow the robot to see the fruit in 3D, enabling it to identify and pick the correct berry. The LED lights enable the robot to operate and pick fruit even in low-light conditions, such as at night.

How does the robot determine whether a picked berry should be considered waste or ready for retail?

After picking a berry, the robot suspends it in an inspection chamber with cameras and consistent lighting to view the berry from all angles. It can detect up to 17 different types of defects, which helps it decide whether to discard the berry or place it in a punnet for retail.

What is the current speed of the robots compared to human pickers?

The current generation of robots is about a quarter of the speed of an average human picker. However, they can operate for multiple shifts, allowing them to pick as much fruit as a person can in a day.

How do the robots avoid bruising the strawberries during the picking process?

The robots avoid bruising the strawberries by picking from the stem rather than grabbing the berry itself. This technique also helps prevent the spread of fungus or other pests.

What is the potential impact of robotic automation on the labor market in farming?

Robotic automation could help address labor shortages and increase efficiency in farming. However, it also poses a threat of displacing human workers, potentially leading to a significant labor disruption if not managed correctly.

How does the robot's design allow it to work in different environments, such as flat farms and steep hillsides?

The robot is designed to be super rugged with super rugged tracks, enabling it to operate in various terrains, from flat farms to steep hillsides, even in challenging weather conditions.

What is the purpose of the robots being designed to work out of shipping containers?

The robots are designed to work out of shipping containers to facilitate their transportation and use in different geographical locations. This allows for more efficient and environmentally friendly operation, as it is cheaper to ship robots than human workers.

How does the data collected by the robots contribute to improving farming practices?

The robots collect about 40 gigabytes of data every day, which can be integrated with weather forecasting to create a yield prediction model. This helps in predicting not just what's ready to pick today, but also what will be ready in the future, leading to more precise and efficient farming practices.

What is the broader vision for the application of robotic intelligence beyond farming?

The intelligence developed for farming robots has the potential to be applied to other dynamic and inconsistent environments like stores, hospitals, or homes. In theory, there's almost no physical job a robot couldn't do, which could lead to a transition in how physical labor is perceived and managed.

Outlines

00:00

🤖 Introduction to Robotic Farming

This paragraph introduces the concept of robot farming, which has been a futuristic idea until recently. It discusses the development of a robot capable of picking soft fruit like strawberries, a task previously only done by humans. The script highlights the current automation on farms and the challenges faced in recruiting human labor for harvesting. The robot's dexterity and intelligence are emphasized, as well as the potential impact of robotic automation on rethinking physical labor. The co-founder of Dogtooth, Duncan, shares his background in machine learning and computer vision, and how his experience in Morocco led to the idea of using technology for environmental cleanup, which later evolved into the development of farming robots.

05:01

🚜 Robotic Innovations in Farming

The second paragraph delves into the design and functionality of the robotic arms developed by Dogtooth. It explains the customization required for the robots to operate in the dynamic farming environment and the various stages of development and testing the robots undergo. The paragraph also describes the technology behind the robot's arms, which include stereo cameras for 3D vision and LED lighting for night operations. The process of picking, inspecting, and sorting the fruit is detailed, along with the current limitations and future improvements of the robots. The role of human workers in supervising and working alongside the robots is also discussed, emphasizing the goal of complementing human capabilities rather than replacing them.

10:05

🌱 The Future of Farming with Robotics

This paragraph explores the broader implications of robotic farming on the agricultural industry and society. It discusses the potential for robots to harvest data and integrate it with weather forecasting to predict crop yields. The adaptability of the robots to work in various environments beyond farming, such as stores and hospitals, is highlighted. The paragraph also addresses the ethical considerations of automation, including the risk of job displacement and the opportunity to create more desirable jobs for humans. It concludes with a reflection on the complexity of tasks that humans find easy and the potential of intelligent robotics to transform physical labor, emphasizing the importance of using this technology responsibly.

Keywords

💡Robot Farmer

A robot farmer is an automated machine designed to perform agricultural tasks traditionally done by humans, such as picking crops. In the video, it is a machine that can pick soft fruit like strawberries, showcasing the advancement in technology and its application in farming to address labor shortages.

💡Farming Automation

Farming automation refers to the use of technology to manage and perform agricultural processes with minimal human intervention. The script mentions GPS-guided tractors as an example of existing automation, but the robot farmer represents a new level of automation capable of delicate tasks like soft fruit harvesting.

💡Dexterity in Robots

Dexterity in robots pertains to their ability to perform precise and complex movements similar to human hands. The video emphasizes the need for robots to be 'smart and dextrous' to keep strawberries intact during picking, highlighting a significant technological advancement in robotic capabilities.

💡Labor Shortage

Labor shortage is a situation where there is not enough workforce available to perform necessary tasks. The video discusses the difficulty in finding human labor to harvest strawberries, leading to significant fruit waste and economic loss, thus motivating the development of robotic solutions.

💡Dogtooth

Dogtooth is the company mentioned in the script that is pioneering the development of the robot farmers. It was co-founded by Duncan, who has a background in machine learning and computer vision, indicating the interdisciplinary nature of the project and its innovative approach to agricultural challenges.

💡Machine Learning

Machine learning is a subset of artificial intelligence that enables systems to learn and improve from experience without being explicitly programmed. In the context of the video, machine learning is crucial for the robots' ability to identify and pick fruit, as well as for future advancements in their capabilities.

💡Computer Vision

Computer vision is a field of artificial intelligence that focuses on enabling machines to understand and interpret visual information from the world. The robots use computer vision to navigate the farm environment, locate fruit, and perform the picking process, which is a complex task requiring the understanding of spatial relationships and object recognition.

💡Fruit Waste

Fruit waste refers to the loss of edible fruit due to various reasons, such as overproduction, lack of labor for harvesting, or damage during handling. The video states that the inability to recruit enough pickers led to approximately 35 million pounds of fruit waste in the U.K., a problem that robotic automation aims to solve.

💡Robotic Arms

Robotic arms are mechanical limbs that replicate the movements of a human arm, allowing for a range of motion and manipulation of objects. The script describes how the robot arms on the farmer are custom-designed with stereo cameras and LED lighting, enabling them to pick fruit with precision and without causing damage.

💡Seasonal Workforce

Seasonal workforce refers to employees who work only during specific periods or seasons, often due to the nature of the job. The video discusses the challenges of securing a seasonal workforce for strawberry picking and how robots can complement human workers during these peak periods.

💡Precision Agriculture

Precision agriculture is an agricultural management technique that uses technology to optimize the application of resources such as water, fertilizers, and labor. The video suggests that robots could enable precision agriculture in new places, potentially reducing food waste and allowing for more efficient farming practices.

Highlights

The development of a robot farmer capable of picking soft fruit like strawberries represents a significant advancement in agricultural automation.

Building a robot with such delicate handling capabilities was previously considered nearly impossible, akin to science fiction.

The U.K. has seen a robot that can navigate in 3D space and approach targets from any angle, which is crucial for handling delicate produce.

There is a significant amount of existing automation on farms, but the new robot represents a leap in technology for soft fruit harvesting.

The smart and dexterous nature of the robots is essential to keep strawberries and similar soft fruits intact during harvesting.

Labor shortages are making it increasingly difficult to harvest all the strawberries, leading to significant economic losses due to fruit waste.

Robotic automation is seen as a potential solution to the labor shortage crisis in agriculture, which is causing millions in lost revenue.

As robots become more sophisticated, they have the potential to redefine physical labor and its role in society.

The founders of Dogtooth, the company behind the robots, come from diverse backgrounds including machine learning and computer vision, not traditional agriculture.

The idea for the agricultural robot came from the need to solve the problem of litter on beaches, showcasing the adaptability of the technology.

Dogtooth's R&D process involves significant custom design and integration of electronics, mechanics, and software for the robots to function in a farm environment.

The robot arms are custom designed to be highly flexible with six degrees of freedom, allowing for complex movements necessary for fruit picking.

The robots use stereo cameras and LED lighting to pick fruit in 3D and operate effectively even in low-light conditions.

An integrated inspection system enables the robots to identify and sort fruit based on quality, deciding whether it's ready for retail or should be discarded.

The design of the robot allows for it to work alongside human labor, performing dull tasks and enabling humans to focus on more complex tasks.

While the current generation of robots is slower than human pickers, they can operate for longer periods, potentially picking as much fruit over a day.

The robots are designed to be mobile and operate from shipping containers, allowing them to be easily transported and used in different locations, including Australia.

The use of robots in agriculture could lead to a more sustainable and efficient food production system, reducing carbon emissions and food waste.

The data collected by the robots during operation can be used to predict yield and optimize agricultural practices, contributing to a more precise agriculture.

The technology and intelligence developed for agricultural robots could be transferred to other dynamic environments like stores, hospitals, or homes, expanding the potential applications of the technology.

The advent of robotic automation in physical labor presents both opportunities for creating better jobs and risks of displacement, emphasizing the importance of thoughtful implementation.

The development of robots to perform tasks that are easy for humans but complex for machines highlights the remarkable capabilities of human beings.

Transcripts

- This is a robot farmer.

- Until recently, building a robot with the same capability

has been almost science fiction.

- We came to the U.K.

to see a robot that can do something totally new-

it can pick soft fruit like strawberries.

Now, there's already a lot of automation on farms:

GPS-guided tractors, for example,

but nothing quite like this.

- Here we're trying to position

something very delicate and small.

We can move to anywhere in 3D space,

and we can approach that point from any angle.

- If you wanna keep the strawberries intact,

you need something smart and dextrous,

which is what these robots do.

Why do this?

It turns out, it's getting harder and harder

to find the human labor we need

to harvest all the strawberries made

in a place like this huge, fancy strawberry farm.

- The cost of fruit wasted last year in the U.K.,

purely because of our inability to recruit enough pickers,

was about 35 million pounds.

That number is forecast to double again this year.

And for me, this is a catastrophe,

and robotic automation can help enormously.

- And as these robots get smarter, and more dextrous,

they can do more than that.

They can actually help us rethink

everything we know about physical labor.

- It takes a lot of effort, intelligence,

and body-minded perseverance to make it all work.

- This is "Hard Reset,"

a series about rebuilding our world from scratch.

- Farming is probably humanity's oldest industry,

but, increasingly,

we're having to rethink all kinds of things.

- This is Duncan.

He is one of the co-founders of Dogtooth,

and his background is in farming.

- So my background is not in agriculture at all.

My background is in machine learning and computer vision.

- Wow. Well, the idea for Dogtooth

came to Duncan while he was on a farm.

- I was sitting on a beach in Morocco.

- Well, God-damn it.

- Morocco is an incredibly beautiful country,

but it is covered in litter.

And it just struck me, that knowing what I knew

about computer vision and machine learning,

actually the problem of identifying the litter

on the beach using cameras, was largely solved.

What we needed, was a low-cost robot

that could crawl their way around the beach,

and pick up all the pieces of litter.

That underlying thinking

took us in a slightly different direction.

- It took them to a farm.

Hey, third times the charm!

This is Ed, Duncan's co-founder,

and he's gonna show us how the robots get built.

- Welcome to Dogtooth.

This is our R&D workshop.

Our robots are made up of a few different technologies,

if you like, electronics, mechanical and software.

And they're all brought together with bits of firmware,

and this is where most of our integration work happens.

So here you'll see our clean room:

This is where we do all of our electronics integration work.

- In order to make the robot work

in such a dynamic environment like a farm,

the folks at Dogtooth

couldn't use a lot of off-the-shelf technology.

They practically had to make every part of this robot

from scratch.

- So this is our commissioning area.

So on the right here, we do arm commissioning.

- This robot arm:

how much of this is custom designed by you?

- All of it, unfortunately.

This is all of our stuff. Yeah.

These arms, here, are in different stages of commissioning.

The ones without any covers there

are on their initial test.

So once it's done its initial commissioning test,

it gets its covers put on,

and then it will run through a burn-in test,

and that takes about 16 hours.

- So this is like C-3PO in "Phantom Menace,"

and then this is from "A New Hope."

- There you go.

I've never seen "Star Wars."

- What?

- I know the reference, where it's from-

but I don't know "Star Wars," I'm afraid.

- Oh my gosh.

- Like the "Star Wars" movies,

robots are not new.

They've been around for decades,

and there's very little excuse for not having seen them.

But unlike C-3PO, robots in the real world

could only work in structured environments

with very tight tolerances.

In other words, nothing like a farm.

- Is this like driving with a PlayStation joystick?

- Exactly right. Yeah.

So my wasted teenage years weren't so wasted.

- See, mom?

So this is one of our fourth-generation robots.

The nice thing about our arms

is that they are super flexible.

It's got six degrees of freedom,

and at any given joint, the actual ability of the arms

rotate way more than 360 degrees.

In this pickerhead here, you'll see a pair of cameras,

and a bunch of LED's.

The LED's allow us to pick at night.

The cameras allow us to see the fruit in 3D.

And in the center here,

we have all our inspection system,

where, once we pick the berry,

we can suspend it into this chamber,

got cameras so we can view the berry all around,

nice consistent lighting,

pick up 17 different types of defect.

And so we know

whether to put it into the waste chute,

or into the punnet ready for retail.

- What do you guys call 'em?

- Punnets.

- Punnets?

- Yeah.

- You guys have crazy names for things. I love it.

- I think in U.S. they're clamshells. Right?

- That sounds right.

- Yeah. - Yeah.

- Or just the box your strawberries come in.

- Yeah, sure.

- Yeah. So here's our polytunnel.

We put this up about three years ago and we are actually,

it turns out, all right

at growing strawberries for a bunch of engineers.

We're going all right.

- So the first question I have as someone

from the United States, I've never seen a strawberry farm

at arm's height.

- Yeah.

- This right all the way already feels like a better design

just from a picking standpoint for humans.

- Absolutely is.

Here in the U.K., we tend to use quite

modern tabletop growing systems.

Approximately the right height to make it easy

for human workers to pick the fruit.

- Obvious, about the robot, is

it's gotta be super rugged.

We're here on a nice flat farm in the middle

of Cambridgeshire.

We also operate these on very steep hillsides

when it's raining cats and dogs and there's mud everywhere,

so we've got super rugged tracks.

Moving on up, you've got the two robot arms

which are doing all the clever bits.

- The arms of the robot

have these stereo cameras and they use computer vision

to navigate this complex environment,

find a stalk, and gently remove the fruit.

- So welcome to the glasshouse up here.

You'll see us starting to run robots.

We've got six running today, I believe.

What we're essentially doing is just looking for clean lines

on which to pick so that we're maximizing our chance

of getting our target berry.

Once we've got a clear vector to the stalk that

they're interested in, they'll pick along that vector

grab a stalk, grip and cut it.

And then they'll take it to the inspection chamber here.

You'll see a few flashes as it images it all around.

Then it will decide which punnet to put it in.

- Well, it's interesting 'cause on this one

it looks like it picked the right berry,

it just was attached to a couple of the wrong ones.

- You'll see here it just picked a little bit high,

and so it got another berry in with it.

- Right. - Pick that up

an inspection chamber,

and so it's put it into a punnet here for supervisor

to sort through it,

pick off the good berries and leave the bad berries behind.

- Just to be clear,

this robot won't mean the end of humans on farms,

so there will still be plenty

of dating profiles on farmersonly.com.

- The dream, I think, is that robots can perform some

of the dull tasks and allow us to focus

on the things that we as human beings are uniquely good at.

- So here at Dogtooth, the goal isn't

to replace human workers.

But the workforce for something

like strawberry picking, which is a seasonal job,

it's very difficult to make sure

that we are getting enough workers across the U.K.

So the way that we view it is that the robots we

can create will be very good at doing a very specific task.

For us, we look to incorporate our robots with human workers

so the humans can do the thinking that the robot can't.

- For now, humans will supervise teams

of these robots as they harvest berries.

And the robots, for now, are a bit slower

than the humans at picking.

- How long does it typically take for one to finish a row?

- Depends how many fruits you have in that row.

- Okay, fair enough.

- One of the things that we haven't prioritized yet

is making them go faster.

- With these robots, we've gone

for the slow but steady approach.

So kind of maximum speeds that we have around these robots

is about a quarter of the speed of an average person.

The next generation we're working on will be more

like half the speed, maybe even more like three quarters

of a speed of a person.

But because they can run for multiple shifts in the day,

they're able to pick off just much fruit as a person can

over a course of the day.

- Another interesting difference is

that these robots don't grab the berry itself-

they pick from the stem.

This has the benefit of making the robot a little bit

less complex, but mainly it avoids the problem of bruising,

and prevents the spread of fungus or other pests.

- So as someone who buys a lot

of strawberries, how do I know

whether or not it's been picked by a robot?

- On the end of each stalk,

and you can see this one which has just been picked here,

there's a little crimp mark, just below the top-

and that's where it's been held onto by the robot.

- Ironically, these robots were made to

supplement a dwindling migrant workforce,

but now, they themselves are a migrant workforce.

- So these are designed to work out of shipping containers.

We've currently got 16 of these

in their shipping container on their way to Tasmania.

- These robots spend part of the year

in Australia, but it's far more affordable

and environmentally friendly to ship robots

to Australia than humans.

And believe me, no one knows more

about how expensive it is to ship humans to Australia

than the English.

Robotics will make it profitable to grow crops

in more places, closer to where they're consumed,

and that will reduce the carbon impact

and make delicious fresh produce available to more people.

- Yeah, I was talking to some farmers yesterday

about whether this

robotics was a revolution or not

and I was saying, "It's not revolution, it's a transition."

If you look at it now, the difference

between a modern farm and a farm 150 years ago is huge.

Well, that didn't happen overnight.

We're just at the start of really applying this kind of idea

of intelligent robotics to real worlds, tangible problems,

and it's a really exciting journey we're all on.

- Very cool.

- So, picture a scenario where robots are involved

in harvesting all of our produce.

The economics of food production would be totally different.

Robots like this wouldn't just address labor shortages:

They'd address food shortages by allowing precision

agriculture in places where it just wasn't possible before.

I don't think most people realize how big of a problem

food waste is.

It's 40% in the U.K. We throw 40% away of fresh produce.

It's nuts, isn't it?

So each robot is also gonna be harvesting

about 40 gigabytes of data every day.

We're now working at building that together

with weather forecasting into a yield prediction model

where we can tell the future

and understand not just what's ready to pick today,

but what will be ready in two weeks time.

- And the intelligence that allows a robot

to work in an inconsistent environment like a farm

will eventually develop into the kind

of intelligence that enables them to work in other dynamic

inconsistent places like stores, hospitals, or homes.

In theory, there's almost no physical job

a robot couldn't do.

So, where does that leave humans?

- I think the hope is that automation will allow us to

create more desirable jobs for people.

But I think there is also a threat.

The danger I see is

that we use robotic automation to displace people,

to rob people of the opportunity to earn a livelihood.

- Intelligence applied to robots

has the potential to completely change everything

we know about physical labor.

If we do it wrong it could cause

the greatest labor disruption in human history.

But if we do it right, it will free

up an even greater source of adaptable intelligence

with the capacity to shape the world:

Us.

- I think the interesting observation is

that we're teaching robots to perform functions

that we as human beings find really easy.

As people, we are remarkably good

at performing what are actually incredibly complex tasks.

It should make us celebrate the beauty

and capability of humanity.

- I'm from America. What's a kilo?

- A kilo is half a pound. 2.2 pounds to a kilo.

- I only use metric system when I'm buying drugs.

Never bought a kilo of drugs-just to be clear.